Knowledge Graph Construction And Application Of Research Literature In The Field Of Entrepreneurship

With the development of knowledge graph technology,many excellent knowledge graph platforms have been built,such as Aminer platform or Microsoft academic graph,those aimed at integration of different academic data and provide better academic searching experience.But these academic knowledge graphs are designed oriented to general domain and on the extraction of entities only papers,journals,academics and other general-purpose entities are considered meanwhile lots of semantic information have not been excavated,the essential problem of the lack of scientific definition and analysis of the literature entity depth,this paper defines the semantic entities implied in the literature,by using Text CNN classification algorithm to associate research methods with paper entities,integration of semantic information by means of a random forest algorithms sort out its semantic pattern,construct a multi-semantic attribute entity model,build a multi-dimensional knowledge graph based on semantic information,provide researchers with multi-dimensional semantic retrieval methods to help them quickly understand the content of the literature.This paper selects the literature in the field of entrepreneurship to analyze,research in this field has arisen since around 1980 and has quickly entered the research peak while being favored by research scholars,the content of the literature is detailed,the characteristics of the field are obvious,and the number of papers is in a stage of rapid growth,existing generic entity types are no longer well suited for multi-dimensional information searches for data.This paper makes use of knowledge graph technology,constructing conceptual entities with reference to the semantic information of common and important research methods in the field of entrepreneurship and use Text CNN to classify and locate text at the same time,the semantic mode of this entity in the abstract is marked and automatically extracted,construct a semantic knowledge graph based on the entrepreneurial field,and provide scholars with multi-attribute semantic search services including research methods.When a scholar searches for a certain type of research method,by using semantic information to determine whether if the paper contain targeted research method,highlighting the relevant sentence simultaneously to help scholar to further understand the information of the paper,thereby achieving rapid positioning and understanding of the retrieved content.The main contents of this paper are as follows:(1)Define the data schema of knowledge graph in entrepreneurial field,including the relationship between entity categories,attributes and entities.(2)Apply convolutional neural network Text CNN algorithm to implement the atomic classification based on papers abstract and title for accomplish the task of entity identification of research method,through the experiment about comparison with other classification algorithms,the algorithm which selected by this article is more accurate and has better effect on recall rate.(3)Annotate semantic entities by semantic mode,by using two-classification algorithm of integrated semantic features to achieve an automated extraction for sentences within target paper those include description about research methods.(4)Design and implement a multi-semantic information document retrieval system around the entrepreneurial field through knowledge storage and query.Experiment shows that,accuracy and recall rate of the automatic classification algorithm and the sentence extraction algorithm those have been used by this article has better performance,therefore ensured the construction of knowledge graph with better robustness.Meanwhile the realization of the multi-semantic information retrieval system in the entrepreneurial field not only helps researchers to quickly find relevant documents,but also helps to understand the literature information from multiple perspectives,so that in today's surge of information volume,the hidden information of data can still be fully exploited and efficient use and dissemination of knowledge.